Cleanup won’t affect homepage submissions: When you clean up or delete media submitted to the homepage from the storage page, your submissions will remain visible on the homepage and won’t be removed.
2026-10-27
Storage & Trash: Open it under “Data” in your account center. Browse thumbnails to choose which assets to clean up. Images currently in use on the canvas won’t be cleaned up by default. Deleted and cleaned-up assets are first moved to Trash and can be restored within 30 days. Space is only freed after you empty the Trash.
Organize Canvas in One Click: Click “Organize Canvas” in the bottom-left canvas toolbar (shortcut: Shift+A) to arrange images from left to right based on their reference relationships, with unlinked images arranged in a grid. When multiple images are selected, only the selected images are organized. Use Undo to restore the original layout.
Results Appear Beside Reference Images: When you generate using an image on the canvas as a reference, the result appears beside that reference image and is connected to it with a line. If the reference image isn’t in the current view, the result still appears in the center of the view.
2026-10-26
Personal data export has been removed: The account page no longer offers personal data export. If you need to review or copy your personal information, please contact us through a support ticket.
2026-10-22
Remember your last model and settings: Previously, the image creation box on the homepage and newly created video boxes on the canvas always opened with the first model. Now, the homepage, workspace, and canvas all open with the model you used last time, along with its size, quality, and number of images.
2026-10-21
Submit to Homepage: Select an image or video you generated on the canvas, then click “Submit to Homepage” in the “⋯” menu to submit it. You can choose whether to make your prompt public. You’ll receive an in-app notification with the review result. View and withdraw your submissions under “My Submissions” in your profile.
2026-10-20
Pick up your last conversation when you open a project: Previously, you had to manually switch to the relevant conversation in your history to see videos you had submitted. Now, opening a project automatically takes you to its most recently used conversation and restores the creation mode you were using.
Play videos on the canvas: Hover over a video card on the canvas to reveal playback controls. Click to play / pause, drag to seek, or toggle sound on and off.
Model pricing: “Model pricing” in the upper-right corner of the credit packages page lists the credits used by each image and video model, along with the equivalent cost (as low as a specified price per image or per second).
2026-10-18
Video prompt optimization: When optimizing video prompts, rewrite them according to the official format for the selected model (Seedance, MiniMax H3, Wanxiang 3.0), taking the duration and added assets into account.
Asset reference corrections: Previously, after deleting and re-adding the first frame or switching video models, references such as "@Image 1" in prompts could point to different assets. They now update automatically with the assets.
2026-10-16
Seedream layer edits preserve transparency: Use Seedream 5.0 to edit extracted layers or images with the background removed, and the results retain a transparent background. Splitting layers from images in formats such as WebP or from smaller images previously failed; it now works correctly.
万相 widescreen aspect ratios: Previously, the 万相 model could generate square images when widescreen ratios such as 21:9 were selected. It now generates images at the selected ratio.
Model Status location: “Model Status” has moved from the left navigation to the “More” menu in the bottom-left corner.
Skills Marketplace: “Skills” has been added to the left navigation. Browse skills shared by other users, add them to use directly from the Agent’s skills, or clone a copy to customize. You can also share your own skills to the marketplace. Each submission requires at least one example image and is listed after admin approval. Must be enabled for the site.
2026-10-15
Create skills in a chat: In an Agent chat, say “Help me create a Skill,” or after you’re happy with a creation, say “Save what we just made as a Skill.” After asking a few clarifying questions, the Agent will save it directly. On the “My Skills” page, you can edit how a skill works and enable or disable it.
Skill example and reference images: Skills can include example and reference images. When you use a skill to generate images, it will refer to these images. Click “Details” in the skill picker to browse the example images.
Updated profile center: It’s divided into four sections: Overview, Credits & Orders, Account & Security, and Preferences & Data. You can change your password and link your email and WeChat account under “Account & Security.” Accounts that can only log in with WeChat will see a prompt recommending that they link an email address and set a password.
2026-10-12
Create images with transparent backgrounds on the canvas: The canvas’s “Image Creation” box previously couldn’t generate images with transparent backgrounds. Now, just turn on “Transparent Background” in its generation settings. The selected model must support this feature.
Support tickets: If you run into an issue, submit a ticket via “More → Support Tickets.” You can paste screenshots directly, and you’ll receive a notification and email when customer support replies. This feature must be enabled for the site.
Articles: The User Guides, Use Cases, and Comparisons & Selection options in the “More” menu at the bottom left have been combined into “Articles.” All tutorials can now be viewed on one page by category.
Help Center: Questions are organized by category on one page. Click a question to view the answer directly.
2026-10-11
Invitation Posters: On the Invite & Earn page, you can save a shareable poster with your referral code and QR code, then post it in group chats or your social feed.
Team Credits: Complimentary credits from trials, check-ins, referral rewards, and similar promotions can no longer be added to a team and can only be used by you.
Contact Us: Open the Contact Us page from the More menu on the left or the Help Center to find contact options such as the support email address and WeChat/QQ group QR codes. You can copy email addresses and account IDs with one click. This feature must be enabled for the site.
2026-10-10
Transparent Background Generation: Turn on “Transparent Background” in the generation settings to create images with transparent backgrounds, ideal for stickers, icons, and product images. Requires model support.
Remove Background on the Canvas: Select an image on the canvas and click “Remove Background” to generate a new image with a transparent background beside it. You can choose 1K / 2K / 4K resolution. Details of the subject may vary slightly.
Thinking Intensity: Some Gemini models let you choose “Fast / Balanced / Deep” in the generation settings. Deeper thinking takes longer to generate images but produces more accurate text and complex layouts. It doesn’t affect credits and must be enabled by the site.
Credit Details: In your account center, you can see how many permanent and expiring credits you have and when each batch expires. Credits are deducted from the batch expiring soonest first. Credits returned after a failed generation retain their original expiration date; previously, they were always converted into permanent credits.
Shared Team Credits: After joining a team, you can switch to using team credits from the credits menu in the top bar. Team administrators can set a monthly allowance for each member. The team feature must be enabled by the site.
2026-10-06
Automatically recommend the number of layers to split: When you select “Auto” for layer splitting, Bihua AI first analyzes the image and recommends how many layers to split it into, while showing the required credits. After you confirm, it splits the image using the recommended number of layers. You can also switch to “指定” and adjust the number yourself.
Agent automatically selects models: Turn on “Auto” in model preferences, and the Agent will choose the right image and video generation models from all available models for each task. You can also select models manually to limit its choices to those models.
Drag images onto the canvas and upload multiple images at once: You can drag images directly from your computer onto the canvas. The upload button and paste function also support uploading up to 20 images at once, which will be arranged on the canvas together after uploading. Drag an image into the input box on the right to use it as a reference image. In video mode, it will be added to the corresponding media slot according to the current method.
2026-10-05
Agent searches online for reference images: When drawing specific real-world things like a particular car model or landmark, Agent searches online for reference images before creating the image. You can see which online images it referenced below the generated image card. Web search must be enabled; no credits are deducted.
2026-10-01
Subscription management: At the top of the credits center, view your current subscription, next charge, and credits used this billing period. You can cancel or resume automatic renewal, change your payment card, and download invoices.
2026-09-30
Fixed inaccessible buttons for cropping and other tools on mobile: Previously, the crop toolbar could extend beyond the screen on mobile. When an image was near the edge of the screen, the upscale, split layers, and reverse prompt panels could also extend off-screen, making their confirm buttons inaccessible. All of these now stay fully within the canvas.
2026-09-29
Move theme and language settings to the profile menu: After signing in, switch themes and interface languages from the profile menu in the top-right corner. On mobile, use the sidebar menu.
Fix an issue where lower models couldn’t be selected when there were many models: Previously, the bottom of the model list could extend beyond the screen when there were many available models or the window was short. The full list now displays within the screen.
Fix image expansion on mobile: Previously, the image expansion toolbar extended beyond the screen on mobile, making the confirm button inaccessible. The toolbar now displays fully across two rows.